Seed Threshold

Node Icon

Create mask regions that are connected to initial seeds within the threshold of \(\mu \pm c\cdot\sigma\), where \(\mu\) and \(\sigma\) are the mean and standard deviation of the selected region, and \(c\) is the multiplier.

The Adaptive Seed Threshold algorithm uses simple statistics to determine which pixels are included in a region. First, the mean and standard deviation of intensity values for all the pixels in the region is computed. A user-provided factor is then used to multiply this value and define a range around the mean. Pixels that fall within this range are accepted as part of the region and considered during subsequent iterations of the algorithm. Once no more neighbor pixels meet these criteria, it's determined that iteration 1 has finished and processing resumes with new parameters. At each stage, updated means and standard deviations are calculated based on all currently included pixels, and the next iteration starts.

Inputs

Input

Input Image(s) to be segmented.

Type: Image, List, Required, Single

Outputs

Output

Segmentation result as Mask(s).

Type: Mask, List

Settings

Seed Type Selection

Define seeds using indeces or positions.

Values: Index [px], Position [mm]

Seed Position [mm] Array

Set the seed.

Seed Index [px] Array

Set the seed.

Number of Iterations Integer

The number of iterations is based on the degree of homogeneity within an anatomical region to be segmented. Highly uniform regions will require a few iterations while more complex regions may need several more. It's important to carefully select the multiplier factor, but there is no guarantee that this algorithm will converge onto a stable region. In practice, it seems most effective to limit the number of iterations in order to avoid potentially segmenting the entire image.

Multiplier Float

Set \(c\) in the condition threshold.

Initial Neighborhood Radius Integer

Set the region to estimate statistics.

Replace Value Float

Set the replace value.